Config_snd_hda_intel_hdmi_silent_stream
by BuckyKatt from LinuxQuestions.org on (#6QFTT)
So, I am still trying to figure out my inconsistent problems related to my Intel ARC A770 video card.
To quickly recap, I seem to have two issues with my A770 ARC... performance is not always great... and sometimes sound over the HDMI is dropping. It seems two things are related to the performance... I cannot enable Resizable BAR... it is not available in my BIOS, and Linux cannot adjust it at boot because there is a PCI bridge in the way. I have a patch for this, but the person disappeared and never explained how to use it. Additionally, in either 6.9.2 or 6.9.3, something was modified in the kernel which causes the A770 ARC to freeze under serious strain. I have gotten around this by only have 200 Firefox windows open, rather than 300. This gives me the ability to boot 6.9 and 6.10 kernels, but since I know the problem isn't actually fixed, I usually revert to a 6.6.x kernel.
The real showstopper is the sound dropping out. I believe the issue is the sound driver for the ARC A770, snd_hda_intel. The machine also has an on-board sound card, which I have hooked to a second set of junk speakers... which never drops out. In the past, I had pretty consistent sound without drops, unfortunately, as newer kernels, BIOS firmware, BIOS settings, A770 firmware have come, the sound has gotten worse. I have swapped between Pulse and WirePlumber countless times. At this point, the sound had dropped to about 95% silence. It wouldn't surprise me if the sound dropping is linked to the problems in the precedeing paragraph.
Enter the CONFIG_SND_HDA_INTEL_HDMI_SILENT_STREAM kernel option... this is the latest rabbit hole I've been down. Reputably setting this to Y enables some keep-alive communication that upsets the A770. Supposedly, this is default N in the stock kernel... and in many popular distros... but in Arch and Slackware this is defaulted to Y.
So, of course I compiled a 6.6.47 kernel without it... and sound seems to drop less... maybe only 20% of the time (much better than 95%)... but that is a seat-of-the-pants test for just a short time.
Is there any particular reason we are using that kernel option?
Also, any ideas for diagnosing the sound path once the snd_hda_intel driver gets ahold of it?
BK
To quickly recap, I seem to have two issues with my A770 ARC... performance is not always great... and sometimes sound over the HDMI is dropping. It seems two things are related to the performance... I cannot enable Resizable BAR... it is not available in my BIOS, and Linux cannot adjust it at boot because there is a PCI bridge in the way. I have a patch for this, but the person disappeared and never explained how to use it. Additionally, in either 6.9.2 or 6.9.3, something was modified in the kernel which causes the A770 ARC to freeze under serious strain. I have gotten around this by only have 200 Firefox windows open, rather than 300. This gives me the ability to boot 6.9 and 6.10 kernels, but since I know the problem isn't actually fixed, I usually revert to a 6.6.x kernel.
The real showstopper is the sound dropping out. I believe the issue is the sound driver for the ARC A770, snd_hda_intel. The machine also has an on-board sound card, which I have hooked to a second set of junk speakers... which never drops out. In the past, I had pretty consistent sound without drops, unfortunately, as newer kernels, BIOS firmware, BIOS settings, A770 firmware have come, the sound has gotten worse. I have swapped between Pulse and WirePlumber countless times. At this point, the sound had dropped to about 95% silence. It wouldn't surprise me if the sound dropping is linked to the problems in the precedeing paragraph.
Enter the CONFIG_SND_HDA_INTEL_HDMI_SILENT_STREAM kernel option... this is the latest rabbit hole I've been down. Reputably setting this to Y enables some keep-alive communication that upsets the A770. Supposedly, this is default N in the stock kernel... and in many popular distros... but in Arch and Slackware this is defaulted to Y.
So, of course I compiled a 6.6.47 kernel without it... and sound seems to drop less... maybe only 20% of the time (much better than 95%)... but that is a seat-of-the-pants test for just a short time.
Is there any particular reason we are using that kernel option?
Also, any ideas for diagnosing the sound path once the snd_hda_intel driver gets ahold of it?
BK